NASA Technical Reports Server (NTRS) 19940023159: A study of DC-DC converters with MCT's for arcjet power supplies

82 Pages·2.9 MB·English

by  NASA Technical Reports Server (NTRS)

/d-2_ o
NASA Contractor Report 191204
A Study of DC-DC Converters With MCT's
for Arcjet Power Supplies
Thomas A. Stuart
University of Toledo
Toledo, Ohio
January 1994
Prepared for
Lewis Research Center
Under Grant NAG3-1102
N96-21662
(NASA-CR-I91204) A STUDY OF OC-DC
CONVERTERS WITH NCT'S FOR ARCJET
POWER SUPPLIES Final Report
Unclas
(Toledo Univ.) 76 p
National Aeronautics and
Space Administration 0206653

Table of Contents
INTRODUCTION ................................................................... 1
I°
A. MCTs vs. FETs ................................................................ 2
B. Resonant vs. Soft Switching Converters .............................. 3
C. Series Resonant Converters ............................................... 5
II. POWER CIRCUIT OF THE 10-KW ARCJET
POWER CONVERTER ............................................................. 9
A. Series Resonant Converter (SRC) Description ...................... 9
B. Design of SRC ................................................................. 17
III. POWER CIRCUIT PERFORMANCE AND
EXPERIMENTAL DATA ................................................. ...... 37
A. Power Circuit Description ............................................... 37
B. Steady-State Performance ................................................ 42
C. Load Transient Tests ...................................................... 51
IV. IGNITION CIRCUIT ............................................................. 59
A. Power Circuit Design ....................................................... 59
B. Experimental Results ...................................................... 64
V. SUMMARY ........................................................................... 73
REFERENCES ....................................................................... 74
Abstract
A Study of DC-DC Converters with MCT's
for Arcjet Power Supplies
by
Thomas A. Stuart
Many arcjet DC power supplies use PWM full bridge converters with large arrays
of parallel FETs. This report investigates an alternative supply using a
variable frequency series resonant converter with small arrays of parallel MCTs.
The reasons for this approach are to: I) Increase reliability by reducing the
number of switching devices; 2) Decrease the surface mounting area of the
switching arrays. The variable frequency series resonant approach is used
because the relatively slow switching speed of the MCT precludes the use of PWM.
The 10kW converter operated satisfactorily with an efficiency of over 91%. Test
results indicate this efficiency could be increased further by additional
optimization of the series resonant inductor.
ii
I. INTRODUCTION
Many of the present DC power supplies for arc jet thrusters utilize a full
bridge DC-DC converter similar to those in [1-3]. These circuits employ four
switches that usually consist of multiple FETs connected in parallel, and phase
shift control is used to regulate the output. The operating frequency is usually
limited to the 10-20 kHz range where switching loss is very low because of the fast
switching times of the FETs. Higher frequencies have been avoided because large
ferrite transformer cores have not been considered satisfactory for a spacecraf%
environment. Most of these converters have been intended for lower power
applications (e.g., 1-2 kW) which are supplied from a 28 Vdc source. At this
voltage, no other switching device is capable of providing the efficiency and cost
effectiveness of FETs.
However, for higher power levels such as 10 kW, it becomes desirable to
increase the source voltage to 150 - 300 Vdc in order to avoid the exceptionally high
currents that would accompany a 28 Vdc source. At this higher voltage level,
other switching devices such as the IGBT and the MCT become competitive with
the FET. This is because the forward voltage drop of the FET climbs rapidly for
breakdown voltages above 200 volts, and a very large number of devices must be
operated in parallel to achieve the forward drop of an IGBT or an MCT. If the
application is not cost sensitive, it is still feasible to operate FETs in this manner,
but very large arrays of FETs are required. This is especially true when
attempting to match the forward drop of an NICT, which promises to have a much
lower forward drop than an IGBT with the same die size [4]. Because MCTs offer
the prospect of a large reduction in the number of switching devices, it was
decided to investigate the use of these devices in a 10 kW arcjet power supply.
1
A. MCTs vs. FETs
The MCT module described in [4] consisted of 6 parallel die and had a
nominal current rating of 300 amps. This device was a prototype MCT power
module developed by the General Electric Corporate Research and Development
Center.
A 10 kW converter with an input voltage of 150 Vdc (specified input voltage
when research proposal was submitted) and an efficiency of 90% will draw an
average input current of 74 Adc. At 74 amps and Tj = 125°C, the GE MCT has a
forward drop of about 0.75 volts. No breakdown voltage ratings were given in [4]
for this device, but it is presumed to be about 600 volts Since it was operated from a
270 Vdc bus and compared with a 600 volt IGBT.
To achieve this same drop at 74 amps with FETs, one might consider the
IRFK4H450 module, which has an Rdson = 0.19 _ at Tj = 125°C and consists of 4
parallel die. This device has a peak voltage rating of 500 volts. To get a drop of .75
V. @ 74 A would require 19 parallel modules, or 76 parallel die. This means that
a full bridge converter would need a total of 4 x 76 = 304 die. While it is feasible to
construct such a circuit, it should be remembered that a short circuit failure of
any single die will destroy the entire power supply unless each die has an
individual fuse. Packaging restrictions may be another problem, and the drive
circuits will need a high impulse current capability because of the high input
capacitance of the FET arrays. All of these problems would be greatly alleviated
by using MCTs, since a circuit with the same forward drop would contain only 6 x
4 = 24 die.
MCTs are much slower than FETs however, so the MCT would not be
competitive in a conventional 20 kHz phase shift circuit because of its higher
switching loss. The GE MCT in [4], for example, has a current fall time of 1714
n.s. as compared to 70 n.s. for the IRFK4H450. Therefore the MCT would have to
2
be used in some type of resonant or soft switching circuit where switching losses
are reduced.
B. Resonant v_ Soft Switching Convertexs
When this project was first initiated, it appeared that one of the better
candidate circuits might be the ZVS/ZCS (zero voltage switching, zero current
switching) converter shown in Fig. I-1. Earlier versions of this circuit are
described in [3,5-8]. These earlier circuitsare satisfactory for power levels in the
1-2 kW range, but they are difficultto implement at higher power levels. The
more recent version in Fig. I-I overcomes this limitation by using an active
snubber which recycles part of the energy back to the source. To date, a 10 kW 20
kHz version of the circuit has achieved an efficiency of 91.3%, even with high
voltage drop IGBTs and a relativelylow output voltage of36 Vdc [9].
As seen from the timing diagram in Fig. I-1, on a typical half cycle for Q1
and Q4, both devices turn on simultaneously, but Q4 turns off first. When Q4
turns off,C1 provides a path for the load current and vcl rises at a gradual rate.
Thus C1 provides zero voltage switching (ZVS) for Q4. Assuming Q2 has reverse
blocking capability, vcl continues to increase beyond Vs, but when vc1 --VBI, £)5
conducts and vCl isclamped at this value. Since VBI >> Vs, ii israpidly driven to
zero, and zero current switching (ZCS) is achieved for QI since il = 0 before Q1 is
turned off. After ilreaches zero, a very slight reverse current flows through Q1
before itrecovers. D6 isrequired to clamp the voltage across Q3 when Q1 recovers.
Although the circuit in Fig. I-1 has proven to be advantageous for higher
input voltages, two developments occurred which indicated that itprobably would
not be competitive for the present application. First of all,the input voltage
specification decreased from 150 to 120 Vdc, and there was some indication that it
might drop even further. The second was that the available MCTs had lower than
anticipated reverse breakdown voltages, and this would have required the use of
, 3
+Vs
o__ _._ o_ _ _o
I LII._/P- _ I "r----b_-,-rY_m-_--_÷
I | | L(}: leokoge inductance oF TI
__J_
' •+¥s I
,4-
C4 --
X_o-
I D6N
C3=
'BI=VB2>Vs VBlRegutotor, AI ._j
Active snubbers
(o. } Power circuit with octive snubbers.
I : I I :
l]4
I I "
Q2
I I I
Ol
03
I I
(b.) tining diocjron For gore drivers.
F icjTj-I. ZVS/ZCS converter with PiCT's
4
series blocking diodes D1-4, as shown in Fig. I-2. The lower input voltage plus the
extra drop of the diodes indicated that the ZVS/ZCS probably would have an
excessive conduction loss as compared to other topologies. Therefore, it was
decided to use the conventional series resonant circuit in Fig. I-3 instead of the
ZVS/ZCS.
C. Series Resonant Converters
The series resonant circuit (SRC) in Fig. I-3 is well known in the power
electronics community, and it has been analyzed extensively [10-17]. If slower
devices such as IGBTs or MCTs are used for Q1-Q4, the circuit should be
controlled by varying the frequency (instead of PWM) in order to reduce the turn-
off switching losses. D5 and D6 also have very low turn-off loss, but D1-D4 still
have significant turn-off loss. All of the conduction losses will be higher than for
conventional phase shift PWM because the current has a higher RMS/average
ratio. This indicates an increased disadvantage for FETs in this circuit because of
their I2R loss characteristic. In spite of the advantages for Q1-Q4 and D5-D6, it
also should be noted that the higher AC currents are a disadvanl;age for all of the
power level capacitors and magnetics because of their higher losses.
Although the SRC exhibits higher conduction losses, the 15 kHz MCT version
achieved an efficiency of 91% at 10 kW. However, it should be noted that this was
accomplished only after an extensive re-design of T1 and Lr, and the temperature
rise of Lr in particular was still higher than expected. This proved to be the case
even when foil conductors or cables with #22 magnet wire were used (although
the individual strands in this cable were insulated, it would not be proper to call
this a Litz cable since the strands were not interwoven). Since these conductors
appear to be satisfactory for the skin depth at 15 kHz, the excessive heating
appears to be related to proximity effect losses and possibly heat transfer problems
associated with larger cables. These heating problems did not appear to be caused
5
+Vs
D9 Lo
Vo
4- T
VCI
04 DIO
03_XD2 __
L O.= leokoge Inductonce oF Tt
J___
I-- _ --
VB2Recju_otor, A2 -=--7
n +¥$
C4! +-VB2
X2 xl
I
ID s
4-
C3: = V81
I ---- : o__.o+v. I
4,
VBlRegulotor. AI
VBt=VB2>Vs
Actlve snubbers
Fig I-2. ZVSIZCS converter with MCT's and blocking diodes.
6